The cloud brought us the age of Big Data and the Internet of Things (IoT). We solved the compute and storage problems (became cheap) and managed to scale resources at a snap of a finger or with a couple lines of code. We can now store unmeasurable quantities of information and make educated decisions based on that…or can we?
The purpose of hoarding all this data is to have multiple angles that would allow us to improve business performance and find weak spots. I guess we get so excited with all of this readily available information we lose track of the end goal. There are so many platforms that offer hundreds of reports that look beautiful and seem sharp, but when we try to trim it down to what we need, we still don’t have that piece of information that allows us to act.
I am also guilty as charged. Not so long ago, I was involved in a project that required collecting quite an amount of data and process it through Hadoop. Well, when the team and I went to design the reports, boy did they look beautiful! We had hundreds of graphic reports from bubbles, heat maps, stacked bars with trend lines, you name it, all lined up perfectly in a stunning dashboard. Then someone asked an interesting question…“How can we use this? How will it help our customers increase their revenue or cut costs?” I remember almost being offended by that question. I thought, “how can you not get the needed information from this beautiful… oh wait you are right!” It’s true, it looked dazzling but by taking a glance I could not make a proper informed decision that would help my client’s business. When we converted all that information into actionable data, the overcrowded (but dazzling) dashboard had only a few graphs. The difference was that, with just a quick glance, we could tell exactly what was right and wrong with my client’s operations and what actions were needed.
This is where Artificial Intelligence (AI) and Machine Learning (ML) are taking their righteous place among us. They were designed to help us make smarter decisions or make those decisions for us even if we present them with an overcrowded dashboard. While there are apocalyptic concerns that we are turning the Terminator movie script into reality, the benefits outweigh the risk of Skynet enslaving the human race. I know right know you must be thinking, “I don’t have the capacity to add this type of technology, let alone afford it.” Well, you would be surprised that you don’t need the software to be able to decipher the meaning of our existence to claim you have AI and ML.
Let me demonstrate with a practical example, let’s imagine you are a communication service provider offering Unified Communications as a Service to enterprises and want to be proactive in detecting potential issues that would decrease the revenue from customers. As a start, you can collect data points of the service usage using a metric such as minutes. Afterwards, ML can benchmark the current readings and compare them to previous results to calculate the deviation. This would allow ML to determine the trend and learn that certain deviations (positive or negative) are part of that time of the year, therefore reducing false positives. ML would flag with a high level of accuracy any abnormal decrease in your service usage based on past behaviors. That is when AI takes over and can analyze why and take the subsequent actions which would correct the issue and notify you. Sometimes it could be something as simple as changing a SIP trunk, using a different codec, or increasing the allowed bandwidth. So, by now you are raising your eyebrows and saying, “Miguel, that is not very advanced, it’s just a simple alarm with a threshold that calculates averages and a dumb action script….” Precisely my point, the data format must be just right and be actionable enough that you can do it with an alarm and a dumb action script.
I am still amazed how many communications service providers spend millions in “intelligent” platforms that show beautiful reports but fail to answer the simple questions and take the obvious actions.